MOM6 vertical coordinate

[aekiss]

This issue is just a place to gather general thoughts about hybrid coordinates.

Our initial configurations will use REGRIDDING_COORDINATE_MODE = "ZSTAR" with a KDS75 z* vertical grid but at some point we'll explore other options.

[aekiss]

CESM3 will use HYCOM1: #83 (comment)

[aekiss]

Alan Wallcraft's talk today at the CLIVAR/COMMODORE workshop provided many useful insights into the regridding and remapping options in MOM6, particularly

• HYCOM1 vs HYBGEN
• regrid and remap interpolation schemes (PPM_CW, PPM_H4, P1M_H2)
• HYCOM1_ONLY_IMPROVES
• MAX_LAYER_THICKNESS

He explained that while the cubic PPM_H4 method is suitable for remapping,P1M_H2 (linear) is preferred over PPM_H4 for regridding, because PPM_H4 eventually produces unstable layers. In contrast, PPM_CW is suitable for both regridding and remapping.

"Regrid jitter" can occur in HYCOM1, meaning that each regrid step nearly exactly undoes the effect of the previous regrid. This unnecessary regridding produces excessive spurious diapycnal diffusion. This jitter is mostly eliminated by setting HYCOM1_ONLY_IMPROVES = True

HYBGEN has less open-ocean regridding than HYCOM1, and therefore less spurious diapycnal diffusion, but has issues with thick-thin-thick layer structure reducing vertical resolution, and also requires horizontal interface smoothing. HYBGEN can't be used as the vertical interpolator and isn't yet production-ready.

Seemed to suggest that HYCOM1 with PPM_CW and HYCOM1_ONLY_IMPROVES is a good choice for now, despite still having more diapycnal diffusion than expected for an isopycnal-favouring method. Their 0.08° 75-layer Atlantic test config uses these settings, with REGRID_TIME_SCALE of 6hr.